R Notebook

Project 1: ACS Quality of Life per State, Education and Gender

We want to find a relationship betweeen the Attained Education, State, Gender, and the quality of life of US population.

The main assumption is that quality of life is directly proportional to with Salary [USD/hour] and Commute Time to work [hours].

Load the libraries and data

library(dplyr)
library(readr)
library(DT)
library(plotly)
library(ggplot2)
library(RColorBrewer)
library(d3heatmap)
load.data <- FALSE

if(load.data){
  datatable(head(acs14pusa,50), options = list(scrollX=T, pageLength = 10))
}

Basic information about original data

if(load.data){
  dim(acs14pusa)
  dim(acs14pusb)
}

Select relevant columns to reduce memory usage.

if(load.data){
  relevant.columns <- c("SERIALNO", "ST", "SEX", "AGEP", "SCHL", "INDP", "WKHP", "WAGP", "ESR",
                        "PINCP", "PERNP", "JWMNP")
  acs14pusa.cols <- acs14pusa[,colnames(acs14pusa)%in%relevant.columns]
  acs14pusb.cols <- acs14pusb[,colnames(acs14pusb)%in%relevant.columns]
  
  rm(acs14pusa, acs14pusb)
  gc()
}

Bind part a and b, and get basic information

if(load.data){
  acs14pus <- rbind(acs14pusa.cols, acs14pusb.cols)
  dim(acs14pus)
}

Add state names and abbreviations

Credits to Arnold Chua Lau (Spring 2016).

if(load.data){  
  ST.anno = read_csv("./data/statenames.csv")
  ST.anno = mutate(ST.anno, STabbr=abbr, STname=name)
  
  acs14pus = mutate(acs14pus, STnum = as.numeric(ST))
  acs14pus <- left_join(acs14pus, ST.anno, by = c("STnum" = "code"))
  
  select(sample_n(acs14pus,5), starts_with("ST"))
}

Convert data

acs14pus$JWMNP <- as.numeric(acs14pus$JWMNP)
acs14pus$WAGP <- as.numeric(acs14pus$WAGP)
acs14pus$WKHP <- as.numeric(acs14pus$WKHP)
acs14pus$STabbr <- as.factor(acs14pus$STabbr)
acs14pus$SCHL <- as.integer(acs14pus$SCHL)
acs14pus$WAGEHOUR <- acs14pus$WAGP / acs14pus$WKHP / 52
industry.categories = read_csv("./data/industry_codes.csv")

Parsed with column specification:
cols(
  code = col_integer(),
  industry = col_character()
)

education.categories = read_csv("./data/education_codes.csv")

Parsed with column specification:
cols(
  code = col_integer(),
  education = col_character()
)

acs14pus$SCHL <- as.integer(acs14pus$SCHL)
acs14pus <- left_join(acs14pus, education.categories, by = c("SCHL" = "code"))

Write result to a csv file. So we do not need to build them again (takes 30 min in a laptop)

if(load.data){
  write_csv(x = acs14pus, path = "./output/ss14pus_columns.csv" )
}

Summary statistics (Jaime)

Summary Statistics

summary.mean <- acs14pus %>% group_by(STabbr) %>% summarise(mean(na.omit(JWMNP)),
                                                       mean(na.omit(WAGP)),
                                                       mean(na.omit(WKHP)),
                                                       mean(na.omit(WAGEHOUR))
                                                       )
summary.mean[,-1] <- round(summary.mean[,-1],1)
names(summary.mean) <- c("STabbr", "JWMNP", "WAGP", "WKHP", "WAGEHOUR")
datatable(summary.mean, options = list(scrollX=T, pageLength = length(summary.mean$STabbr)))

BoxPlot - Commute Time

BoxPlot - Wage per Hour (Jaime)

JG: I am not convinced of these data values. DC is far too high!

Heatmap

After US population gets a bacherlor’s degree, there is a tendency tend to commute more than if you did not have the higher level degree

acs14pus <- acs14pus[order(acs14pus$SCHL),]
plot_ly(x = acs14pus$education,
        y = acs14pus$JWMNP,
        type = "box",
        sort = FALSE) %>%
  layout(title = "Commute time by Education Attainment",
         xaxis = list(title ="Education"),
         yaxis = list(title = "Commute time"),
        # width = 1000,
        # height = 700,
         legend = education.categories$education
         )

plot_ly(z = summary.median.2$JWMNP, 
        x = summary.median.2$education,
        y = summary.median.2$industry,
        type = "heatmap",
        colorscale = "Hot") %>%
  layout(title = "Commute time by Education Attainment and Industry",
         xaxis = list(title ="Education"),
         yaxis = list(title = "Industry"),
         width = 1000,
         height = 700
         )

Choropleth maps of US states (Ying)